Univariate Plots Section

In this section, I want to explore the dataset regarding to borrowers created in prosper, the Loan Term they are more likely to select, the prosper rating, the borrowers’ occupation and their annual incomes, listing category,borrower’s interest rate, lender yield, loan amount, borrowers late payment and the investors distributions.

##    12    36    60 
##  1614 87778 24545

There are almost 80% of the borrowers more likely to choose 3-year fixed loan term while they creating the listing.

## [1] "Q1 2011" "Q4 2012"

There are 1614 loan listing with fixed Term 1 year, those listing are created from 2010 to 2013, I am not sure if this part of data is correct or not? as from prosper.com, that only support Fixed term—3 or 5 years.

##                              Other                       Professional 
##                              28617                              13628 
##                Computer Programmer                          Executive 
##                               4478                               4311 
##                            Teacher           Administrative Assistant 
##                               3759                               3688 
##                            Analyst                                    
##                               3602                               3588 
##                 Sales - Commission                     Accountant/CPA 
##                               3446                               3233 
##                           Clerical                     Sales - Retail 
##                               3164                               2797 
##                      Skilled Labor                  Retail Management 
##                               2746                               2602 
##                         Nurse (RN)                       Construction 
##                               2489                               1790 
##                       Truck Driver                            Laborer 
##                               1675                               1595 
##  Police Officer/Correction Officer                      Civil Service 
##                               1578                               1457 
##              Engineer - Mechanical                  Military Enlisted 
##                               1406                               1272 
##            Food Service Management              Engineer - Electrical 
##                               1239                               1125 
##                       Food Service                 Medical Technician 
##                               1123                               1117 
##                           Attorney               Tradesman - Mechanic 
##                               1046                                951 
##                      Social Worker                     Postal Service 
##                                741                                627 
##                          Professor                            Realtor 
##                                557                                543 
##                             Doctor                        Nurse (LPN) 
##                                494                                492 
##                       Nurse's Aide            Tradesman - Electrician 
##                                491                                477 
##                    Waiter/Waitress                            Fireman 
##                                436                                422 
##                          Scientist                   Military Officer 
##                                372                                346 
##                         Bus Driver                          Principal 
##                                316                                312 
##                     Teacher's Aide                         Pharmacist 
##                                276                                257 
## Student - College Graduate Student                        Landscaping 
##                                245                                236 
##                Engineer - Chemical                           Investor 
##                                225                                214 
##                          Architect         Pilot - Private/Commercial 
##                                213                                199 
##                             Clergy           Student - College Senior 
##                                196                                188 
##                         Car Dealer                            Chemist 
##                                180                                145 
##                       Psychologist                          Biologist 
##                                145                                125 
##                          Religious                   Flight Attendant 
##                                124                                123 
##                          Homemaker              Tradesman - Carpenter 
##                                120                                120 
##           Student - College Junior                Tradesman - Plumber 
##                                112                                102 
##        Student - College Sophomore                            Dentist 
##                                 69                                 68 
##         Student - College Freshman        Student - Community College 
##                                 41                                 28 
##                              Judge         Student - Technical School 
##                                 22                                 16

The largest number is Other, it is understandable, because all cases not in above can be Other. If we don’t consider the “Other” case, Professional has the most top count and almost three times as many as the second. The count of borrowers on the top three is professional, Computer Programer, Executive, at the bottom of the list are the students, generally, the students don’t need to borrow money, or they don’t have strong credit rating to create loan request from prosper.

##           A    AA     B     C     D     E    HR 
## 29084 14551  5372 15581 18345 14274  9795  6935

Prosper Ratings, from lowest-risk to highest-risk, are labeled AA, A, B, C, D, E, and HR (“High Risk”), we can see 29084 prosper rating is labeled “NA”,this is understandable since the prosper rating is provided since 2009 after SEC registration, there is no rating labeled to the borrowers before 2009. If we do not consider the “NA”, the prosper rating appears Binomial distribution, 57% of borrowers are labeled middle risk B,C and D. Borrowers’s rating C are the highest frequency up to 22%.

##             $0      $1-24,999      $100,000+ $25,000-49,999 $50,000-74,999 
##            621           7274          17337          32192          31050 
## $75,000-99,999  Not displayed   Not employed 
##          16916           7741            806

The borrower’s income ranges appears Binomial distribution. most of borrowers’ annual income are fall into $25,000 - 75,000. and there are 621 borrowers with zero incomes they got loans, it’s default value if prosper not get the source of borrower’s incomes, or still the borrowers can get loans if they have good credit rating from other path.

##                    Employed     Full-time Not available  Not employed 
##          2255         67322         26355          5347           835 
##         Other     Part-time       Retired Self-employed 
##          3806          1088           795          6134

## [1] 11 15

Further exploring the borrowers whose incomes are zero, find 11 borrowers are Not employed, they may not have incomes? all the loan requests were created on Q3 2007. I guess it should be data issue on the person with zero annual income.

##     1     0     7     2     3     6     4    13    15    18    14    20 
## 58308 16965 10494  7433  7189  2572  2395  1996  1522   885   876   771 
##    19     5    16    11     8    10     9    12    17 
##   768   756   304   217   199    91    85    59    52

Category: The category of the listing that the borrower selected when posting their listing: 0 - Not Available, 1 - Debt Consolidation, 2 - Home Improvement, 3 - Business, 4 - Personal Loan, 5 - Student Use, 6 - Auto, 7- Other, 8 - Baby&Adoption, 9 - Boat, 10 - Cosmetic Procedure, 11 - Engagement Ring, 12 - Green Loans, 13 - Household Expenses, 14 - Large Purchases, 15 - Medical/Dental, 16 - Motorcycle, 17 - RV, 18 - Taxes, 19 - Vacation, 20 - Wedding Loans.

Out of the listing category, three categories comes out as the highest past 10000 loans. There’s N/A and Other categories, so we can’t know for sure the specific category. But one comes out highest which is category 1 - Debt Consolidation, where’s one take out a loan to pay many others. This comes really high with 58308 loans,it could be that many Prosper visitors comes with already have loans, and want to search some loans to pay for it?

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##  0.0000  0.1340  0.1840  0.1928  0.2500  0.4975

The borrower’s rate follow an almost unimodal distribution, with the peak around 0.16. There’s small spike occurs around 0.3.

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
## -0.0100  0.1242  0.1730  0.1827  0.2400  0.4925

It’s very similar to the borrower’s rate, the lender’s yield also follow an almost unimodal distribution, with the peak around 0.14. There’s small spike occurs around 0.3.

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    1000    4000    6500    8337   12000   35000

This is interesting, here we see that the distribution is right skewed, with the peak USD 3500, then some small spikes occur around 10000, 20000, 30000 dollars.

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##  0.0000  0.0000  0.0000  0.5921  0.0000 83.0000     697

Here we can see that there is 75% of borrowers no delinquencie record on prosper peer-to-peer lending marketplace, to my surprise, we found the maximum elinquencie up to 83. It’s really high.

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    1.00    2.00   44.00   80.48  115.00 1189.00

We see that 25% of listing are funded by 1 or 2 investors, most of listing are less than 40 investors.

Univariate Analysis

There are 113937 observations and 81 variables in the raw data set. Due to many variables in this dataset, I only select 15 features from dataset prosperLoanData.

Bivariate Plots Section

## 
##  Pearson's product-moment correlation
## 
## data:  as.numeric(ProsperRating..Alpha.) and BorrowerRate
## t = 917.37, df = 84851, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.9524846 0.9537172
## sample estimates:
##       cor 
## 0.9531049

AA LOWEST risk-> HR HIGHEST risk

We see that the higher risk labeled to borrowers,the higher borrower rate they would cost. And we see that it has a strong positive linear relationship. This is can be explained as the interest rate is the one where the loaner has benefit from the money that he/she loans.

## 
##  Pearson's product-moment correlation
## 
## data:  as.numeric(ProsperRating..Alpha.) and LoanOriginalAmount
## t = -138.17, df = 84851, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.4340341 -0.4230486
## sample estimates:
##        cor 
## -0.4285572

It shows rating E and HR has smaller IQR compared to the rest of the rating, rating B and C has no outliner, rating A,B has a smaller outliner, rating AA has two outline only. With a correlation of -0.43, it shows a negative moderately strong linear relationship, as the risk rating increase, the lower amount would be borrowed from prosper.

The distribution has a bell shape and is right-skewed, we can see that investors prefer to loan to those borrowers holding lower risk rating. This is understandable since loan to those person with higher risk may be higher loss the money, even though the lenders would have higher yield by lending to higher risk case (see the next plot).

## 
##  Pearson's product-moment correlation
## 
## data:  as.numeric(ProsperRating..Alpha.) and LenderYield
## t = 917.52, df = 84851, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.9524993 0.9537315
## sample estimates:
##       cor 
## 0.9531194

We see that it has a strong positive linear relationship between risk rating and lenderYield, that means that if the lenders want to have higher yield, they would take on more risk of lose rate.

The LeanderYield quickly increase in the initial 1-year, then in a stable line after 2006. But the thing that surprised me is there was no LenderYield from Q3 2008 to Q2 2009. Then I searched the following from wikipedia.

On November 24, 2008, the SEC found Prosper to be in violation of the Securities Act of 1933. As a result of these findings, the SEC imposed a cease and desist order on Prosper. Due primarily to the novel nature of the peer-to-peer lending models, the SEC, after review, now treats all peer-to-peer lending transactions as sales of securities and requires that all platforms register with the SEC.

In July 2009, Prosper reopened their website for lending (“investing”) and borrowing after having obtained SEC registration for its loans. Hence, that can explain why there were a continuous gap period on lender yield.

Bivariate Analysis

There is a continuous gap period Prosper was not allowed to run their business until it had obtained SEC registration for its loans. The lender yield is stale in the years.

Multivariate Plots Section

In the previous section we explored there is a strong positive linear correlation relationship between prosper rating and interest rate. here, by adding one more variable Term, we see that for each rating, AA to D, looks the longer loan Term would cause the higher interest rate.

## 
##  Pearson's product-moment correlation
## 
## data:  LoanOriginalAmount and BorrowerRate
## t = -117.58, df = 113940, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.3341283 -0.3237719
## sample estimates:
##        cor 
## -0.3289599

This scatter plot shows the relationship of Loan amount, borrow rate and prosper rating. Rating AA,A and B looks can loan amount up to $35000, rating C less than $25000, rating D less than $15000, rating E less than $10000, and the rating highest risk rating HR less than $5000, seeing the plot on y-axis, the interest rate is increasing from risk rating AA to HR.

We can see investors are more likely to invest in low risk request, not the amount of the loan, the investors on AA rating looks obviously higher than others.

## substr(prosper$LoanOriginationQuarter, 1, 4): Q1 2
## [1] 35000
## -------------------------------------------------------- 
## substr(prosper$LoanOriginationQuarter, 1, 4): Q2 2
## [1] 35000
## -------------------------------------------------------- 
## substr(prosper$LoanOriginationQuarter, 1, 4): Q3 2
## [1] 35000
## -------------------------------------------------------- 
## substr(prosper$LoanOriginationQuarter, 1, 4): Q4 2
## [1] 35000

## prosper$ProsperRating..Alpha.: AA
## [1] 35000
## -------------------------------------------------------- 
## prosper$ProsperRating..Alpha.: A
## [1] 35000
## -------------------------------------------------------- 
## prosper$ProsperRating..Alpha.: B
## [1] 35000
## -------------------------------------------------------- 
## prosper$ProsperRating..Alpha.: C
## [1] 25000
## -------------------------------------------------------- 
## prosper$ProsperRating..Alpha.: D
## [1] 15000
## -------------------------------------------------------- 
## prosper$ProsperRating..Alpha.: E
## [1] 15900
## -------------------------------------------------------- 
## prosper$ProsperRating..Alpha.: HR
## [1] 16800

This plot shows that the maximum loan amount in prosper website is $15000 in initial year 2015, $25000 before Q2 2013 and $35000 after that. There should be no business run from Q3 2008 to Q2 2009 on prosper, after colored by risk rating, we also found there is no risk rating provided to borrowers, all of them are NA. The risk rating is implemented after Prosper reopened from Q2 2009. The rating D, E, HR can borrow maximum amount around $15000, rating C is up to $25000, all rating AA, A, B are up to maximum loan amount $35000.

We can see plot are left skewed. it shows that rating E,HR borrowed less amount, but they have more delinquencies. while the credit rating increasing, the fewer number of delinquencies come out, even though the borrowers labeled lower risk borrowed more money.

Multivariate Analysis

As the loan term and risk rating increase, the interest rate will increase also, there is a strong positive linear correlation relationship between prosper rating and interest rate. if we see the loan amount and risk rating, we found the higher risk rating lead borrowers have to pay more interest rate, and borrowers can have less loan amount also.

Investors are also more likely to invest their money to those borrowers they have lower risk and good credit rating, because they don’t want to lost their money, but still can have yield from this.

Further study the current delinquencies of borrowers , that also explains that high-risk borrowers generally have a higher default rate.

Final Plots and Summary

Reflection

The dataset is from the Prosper loans, it’s America’s first peer-to-peer lending marketplace, where people could loans money by listing it in the website. There are 113937 observations and 81 variables in the dataset.

I extract 15 features and explore loan term distribution, the borrowers’ Occupation, rating, annual incomes, employmentStatus, and what purpose they loan from Prosper. There almost 80% of the borrowers selected 3 years loan term. I am surprise that there are 1614 listing was created from 2010 to 2013 with 1 year loan term, prosper.com doesn’t provide such term for loan. Not sure if prosper opened for this term or what problems caused this data issue. While we seeing Occupation of borrowers, the top 1 is Other, it’s understandable because all not in the website selecting list can be other, the second to third are professional and computer programer. The prosper rating are not provided before Q2 2009, Ratings, from lowest-risk to highest-risk, are labeled AA, A, B, C, D, E, and HR (“High Risk”), the distributions of rating follow an almost unimodal distribution, with the peak of 18345. The borrowers incomes range distributions also looks follow an almost unimodal distribution, we noticed that there are 621 loan list with zero annual income, then I further explored their employment status when they creating list, most of them are in employed status, hence, those zero incomes should be data issue. Finally, I explore the purpose of borrowers loan from prosper, found the highest category is 1 - Debt Consolidation, where’s one take out a loan to pay many others. we may guess it could be that many Prosper visitors comes with already have loans, and want to search some loans to pay for it.

Prosper rating is a very important feature, looking at the statistics and plot, Prosper Rating is strongly correlated with borrower’s interest rate, the number of investors, and the LenderYield. It can be explained that this lower risk rating the borrowers labeled, the lower borrower’s interest rate would have, and will attract more investors also. For lenders, if they would like to fund for those borrowers labeled higher risk rating, the higher yield will have, but on other hand, it will increase the money lost rate.

As I only select 15 features for this dataset, there must be additional features also important to expplore the prosper loan data, we can involve more features regarding to borrowers, lenders or prosper.org to explore more interesting results.